FRM Part I · FRM Exam Part I · Regression Diagnostics
An analyst finds two regressors in a model with a pairwise correlation of 0.95 and considers remedies. Which action is most appropriate to reduce multicollinearity without introducing a new specification problem when both variables are theoretically relevant?
Increasing the sample size or combining the correlated variables into a single composite measure is most appropriate. Both reduce the variance inflation without discarding relevant information. Dropping a theoretically relevant variable risks omitted variable bias, and robust standard errors do not address collinearity.
- ADrop one variable regardless of theory, since this always improves estimates
- BIncrease the sample size or combine the variables into a single composite measureCorrect
- CUse robust standard errors, which eliminate the correlation
- DTransform the dependent variable into first differences
Explanation
Collecting more data lowers coefficient variances, and combining closely related variables into one measure removes the redundancy while keeping the information. Dropping a relevant variable risks omitted variable bias. Robust standard errors address heteroskedasticity, not collinearity, and differencing the dependent variable does not change regressor correlation.
Did you get it right without looking?
One question tells you little. A timed set on Regression Diagnostics shows your real accuracy, how long you take and where you lose marks.
More Regression Diagnostics questions
- A risk analyst's regression of portfolio returns on two highly correlated factors (sample correlation 0.97) is used only to forecast returns…
- An analyst regresses squared residuals from an OLS model on the explanatory variables as part of a Breusch-Pagan test. The auxiliary regress…
- A regression has residual standard error s = 2.0. One observation has residual e = 7.0 and leverage h = 0.19. Using the internally studentiz…
- In a cross-sectional regression of household food spending on household income, the dispersion of the residuals grows steadily as income ris…
- A risk manager estimates a time-series regression by OLS and finds the residuals are positively autocorrelated, while the regressors are not…
- After fitting a multiple regression, an analyst finds that a plot of residuals against a regressor shows a systematic U-shaped pattern. Whic…